InkBench OCR

Pairwise VLM-as-judge OCR benchmark ranking OCR models on InkBench handwriting/document images with Bradley-Terry ELO scores.

5rows
eloprimary metric
2026-05-06sampled

Metadata

ID
inkbench_ocr
Category
Multimodal
Release
Unknown

Metrics

ELO, Win%, Wins, Losses (lower is better), Ties, ELO Low, ELO High

Latest Results

Rows are parsed from the public Hugging Face dataset-server rows API. ELO rankings come from OCR Bench pairwise VLM-as-judge comparisons.

Rank Subject ELO Model Match Provenance Sampled
1 zai-org/GLM-OCR 1706 — Imported 2026-05-06
2 lightonai/LightOnOCR-2-1B 1622 — Imported 2026-05-06
3 deepseek-ai/DeepSeek-OCR 1527 — Imported 2026-05-06
4 FireRedTeam/FireRed-OCR 1382 — Imported 2026-05-06
5 rednote-hilab/dots.ocr 1263 — Imported 2026-05-06